arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2407.16772cs.CVcs.CLcs.LG
收藏

VisMin:视觉最小变化理解

VisMin: Visual Minimal-Change Understanding

Mila - Quebec AI Institute(米拉 - 魁北克人工智能研究所);Université de Montréal(蒙特利尔大学)

Rabiul Awal, Saba Ahmadi, Le Zhang, Aishwarya Agrawal

AI总结:

本文提出VisMin基准,通过最小变化图像-描述对测试VLM的细粒度理解,发现现有模型在空间关系和计数上存在缺陷,并生成训练数据微调CLIP和Idefics2以显著提升性能。

摘要:

对物体、属性以及物体之间关系的细粒度理解对于视觉-语言模型(VLM)至关重要。现有基准主要侧重于评估VLM在给定图像时区分两个非常相似的描述的能力。在本文中,我们引入了一个新的、具有挑战性的基准,称为视觉最小变化理解(VisMin),它要求模型在给定两张图像和两个描述时预测正确的图像-描述匹配。图像对和描述对包含最小变化,即每次仅从以下方面之一发生变化:物体、属性、计数和空间关系。这些变化测试了模型对物体、属性(如颜色、材质、形状)、计数以及物体之间空间关系的理解。我们构建了一个使用大语言模型和扩散模型的自动框架,随后由人工标注者进行严格的4步验证过程。实证实验表明,当前的VLM在理解空间关系和计数能力方面存在显著缺陷。我们还生成了一个大规模的训练数据集来微调CLIP和Idefics2,在基准测试中显示出细粒度理解的显著改进,以及CLIP在通用图像-文本对齐方面的提升。我们在https://vismin.net/发布了所有资源,包括基准、训练数据和微调后的模型检查点。

英文摘要:

Fine-grained understanding of objects, attributes, and relationships between objects is crucial for visual-language models (VLMs). Existing benchmarks primarily focus on evaluating VLMs' capability to distinguish between two very similar captions given an image. In this paper, we introduce a new, challenging benchmark termed Visual Minimal-Change Understanding (VisMin), which requires models to predict the correct image-caption match given two images and two captions. The image pair and caption pair contain minimal changes, i.e., only one aspect changes at a time from among the following: object, attribute, count, and spatial relation. These changes test the models' understanding of objects, attributes (such as color, material, shape), counts, and spatial relationships between objects. We built an automatic framework using large language models and diffusion models, followed by a rigorous 4-step verification process by human annotators. Empirical experiments reveal that current VLMs exhibit notable deficiencies in understanding spatial relationships and counting abilities. We also generate a large-scale training dataset to finetune CLIP and Idefics2, showing significant improvements in fine-grained understanding across benchmarks and in CLIP's general image-text alignment. We release all resources, including the benchmark, training data, and finetuned model checkpoints, at https://vismin.net/.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑